How much (little) are the AI companies making?
The 24-city for my novel Picks and Shovels is in the home stretch! Catch me in LONDON TODAY (July 1) with TRASHFUTURE’s RILEY QUINN and at BLACKWELL’S MANCHESTER TOMORROW(July 2).
If there’s one area where tech has shown a consistent aptitude for innovation, it’s in accounting tricks that make money-losing companies appear wildly profitable. And AI is the greatest innovator of all (when it comes to accounting gimmicks).
Since the dotcom era, tech companies have boasted about giving stuff away but “making it up in volume,” inventing an ever-sweatier collection of shell-games that let them hide the business’s true profit and loss.
The all-time world champeen of this kind of finance fraud is Masayoshi Son, the founder of Softbank, who acts as the bagman for the Saudi royals’ personal investments. Remember last decade when the tech press was all abuzz about “unicorns” — startups that were worth $1b? That was Son: he would take a startup like Wework, declare its brand to be worth $1b, invest an infinitesimal fraction of $1b in the company based on that valuation (sometimes with a rube co-investor) and declare the valuation to be “market-based.” A whole string of garbage companies achieved unicornhood by means of this unbelievably stupid trick:
https://pluralistic.net/2022/05/27/voluntary-carbon-market/#trust-me
Of course, every finance bro is familiar with Stein’s Law: “anything that can’t go on forever eventually stops.” Sure, the Saudi royals could be tapped to piss away $31b on Uber, losing $0.41 on every dollar for 13 years, but eventually they’re going to turn off the money spigot and attempt to flog their shares to retail and institutional suckers. To make that work, they have to invent new accounting tricks, like when Uber “sold” its failing overseas ride-hailing businesses to international rivals in exchange for stock, then declared that these companies’ illiquid stock had skyrocketed in value, tipping Uber into the black:
https://pluralistic.net/2022/08/05/a-lousy-taxi/#a-giant-asterisk
Even companies that are actually profitable (in the sense of bringing in more revenue than it costs to keep the business’s lights on) love to juice their stats, and the worst offenders are the Big Tech companies, who reap a vast commercial reward from creating the illusion that they are continuing to grow, even after they’ve dominated their sector.
Take Google: once the company attained a 90% global search market-share, there were no more immediate prospects for growth. I mean, sure, they could raise a billion new humans to maturity and train them to be Google customers (e.g., the business plan for Google Classroom), but that takes more than a decade, and Google needed growth right away. So the company hatched a plan to make search worse, so that its existing users would have to search multiple times to get the information they sought, and each additional search would give Google another chance to show you an ad:
https://pluralistic.net/2024/04/24/naming-names/#prabhakar-raghavan
But that was small potatoes. What Google — and the rest of the tech sector — needed was a massive growth story, a story about how their companies, worth trillions of dollars, could double or triple in size in the coming years. There’s a kind of reflexive anti-capitalist critique that locates the drive to tell growth stories in ideology: “endless growth is the ideology of a tumor,” right?
But spinning an endless growth story isn’t merely ideological. It’s a firmly materialistic undertaking. Companies that appear to be growing have market caps that are an order of magnitude larger than companies that are considered “mature” and at the end of their growth phase. For every dollar that Ford brings in, the market is willing to spend $8.60 on its stock. For every dollar Tesla brings in, the market is willing to spend $118 on its stock.
That means that when Tesla and Ford compete to buy something — like another company, or the labor of highly sought after technical specialists — Tesla has a nearly unbeatable advantage. Rather than raiding its precious cash reserves to fund its offer, Tesla can offer stock. Ford can only spend as many dollars as it brings in through sales, but Tesla can make more stock, on demand, simply by typing numbers into a spreadsheet.
So when Tesla bids against Ford, Ford has to use dollars, and Tesla can use shares. And even if the acquisition target — a key employee or a startup that’s on the acquisitions market — wants dollars instead of shares, Tesla can stake its shares as collateral for loans at a rate that’s 1,463% better than the rate Ford gets when it collateralizes a loan based on its own equity:
https://pluralistic.net/2025/05/07/rah-rah-rasputin/#credulous-dolts
In other words, if you can tell a convincing growth story, it’s much easier to grow. The corollary, though, is that when a growth company stops growing, when it becomes “mature,” it experiences a massive sell-off of its stock, as its share price plummets to a tenth or less of the old “growth” valuation. That’s why the biggest tech companies in the world have spent the past decade — the decade after they monopolized their sectors and conquered the world — pumping a series of progressively stupider bubbles: metaverse, cryptocurrency, and now, AI.
Tech companies don’t need these ventures to be successful — they just need them to seem to be plausibly successful for long enough to keep the share price high until the next growth story heaves over the horizon. So long as Mister Market thinks tech is a “growth” sector and not a “mature” sector, tech bosses will be able to continue to pay for things with stock rather than cash, and their own stockholdings will continue to be valued at sky-high rates.
That’s why AI is being crammed into absofuckingloutely everything. it’s why the button you used to tap to start a new chat summons up an AI that takes seven taps to banish again — it’s so tech companies can tell Wall Street that people are “using AI” which means that their companies are still part of a growth industry and thus entitled to gigantic price-to-earnings ratios:
https://pluralistic.net/2025/05/02/kpis-off/#principal-agentic-ai-problem
The reality, of course, is that people hate AI. Telling people that your product is “AI enabled” makes less likely to use it:
https://www.tandfonline.com/doi/full/10.1080/19368623.2024.2368040#d1e1096
People — who have had an infinitude of AI crammed into down their throats — are already sick of AI. Policymakers and financiers — credulous dolts who fall for tech marketing hype every! fucking! time — are convinced that AI Is The Future. This presents a dilemma for tech companies, who research the hell out of how people actually use their products and thus must be extremely aware of how hated AI is, but whose leadership is desperate to show investors that they are about to experience explosive growth through the miracle of AI.
The reality is that AI is a very bad business. It has dogshit unit economics. Unlike all the successful tech of the 21st century, each generation of AI is more expensive to make, not cheaper. And unlike the most profitable tech services of this century, AI gets more costly to operate the more users it has.
You can be forgiven for not knowing this, though. As Ed Zitron points out in a long, excellent article about the credulity and impuissance of the tech press, the actual numbers suuuuuck:
https://www.wheresyoured.at/make-fun-of-them/
- Microsoft
Spending: $80b in 2025
Projecting: $13b in 2025
Actually: $10b comes from Openai giving back compute credits Microsoft gave to Openai, bringing the true total to $3b.
- Meta
Spending: $72b in 2025
Receiving: At most $600m in gross revenue from selling “smart” Raybans, which might not actually be loss-leaders, meaning it’s possible that they’re making less than $0.00.
- Amazon
Spending: $100b in 2025
Projecting: $5b in revenue in 2025
Spending: $75b in 2025
Projecting: They won’t say, possibly zero.
As Zitron points out: this industry is projecting $327b in spending this year, with $18b in revenue and zero profits. For comparison: smart watches are a $32b/year industry.
Now, what about Openai? Well, they’re one of Masoyoshi Son’s special children, of a piece with Wework and Uber. Openai is projecting $12.7b in revenue this year, with losses of $14b. Add in a bunch of also-rans like Perplexity and Surge, and the revenue rises to $32.3b. But…if you chuck them in, you also get total exenditure of $370.8b.
These are by no means the only funny numbers in the AI industry. Take “Stargate,” a data-center initiative with a price tag of $500b. Actual funds committed? $40b.
These are terrible numbers, but also, these are some genuinely impressive accounting gimmicks. They are certain to keep the bubble pumping for months or perhaps years, convincing gullible bosses to fire talented employees and replace them with bumbling chatbots that will linger for years or decades, the asbestos in the walls of our high-tech civilization.
If you’d like an essay-formatted version of this post to read or share, here’s a link to it on pluralistic.net, my surveillance-free, ad-free, tracker-free blog:
https://pluralistic.net/2025/06/30/accounting-gaffs/#artificial-income
Image:
Cryteria (modified)
https://commons.wikimedia.org/wiki/File:HAL9000.svg
CC BY 3.0
https://creativecommons.org/licenses/by/3.0/deed.en
